HKIER

摘要

結構方程分析能協助研究者更準確估計一些潛伏變項之間的關係，且能提供一些指標，顯示模式與數據的吻合程度，本文指出在驗証模式時，應留意其他同等吻合模式的存在，此外，並以實例介紹構劃及刪除這些等同模式的方法。文章亦利用省儉原則（principle of parsimony）及窩套模式（nested models）概念，討論如何比較連串模式，從而找出最佳者。

Abstract

Structural equation modelling can give better estimates of the relationships among latent variables, and can also provide reasonably objective indexes on the goodness of fit of models to data. The present article points out the importance of inspecting equivalent models which have the same goodness of fit to the data as the posited model. Rules in constructing equivalent models and ways to eliminate them are discussed with concrete examples. Furthermore, strategies to search for the best model using the concept of nested models and the principal of parsimony are also discussed.